Subtitles: how to get an SRT file with timestamps
The 'Subtitles' tool recognizes speech in an uploaded audio file directly in the browser and creates an SRT file with timestamps. In this article we will explain how to properly prepare the audio, select the recognition language, add context to the text field and check the ready result. You will learn which parameters really affect quality, how to evaluate timestamp accuracy and what limitations are important to consider. The description is based on the real capabilities of the tool, without fictional functions.
Open tool: Subtitle generator ↗How the 'Subtitles' tool works
The 'Subtitles' tool works entirely in the browser: you upload an audio file up to 20 MB or up to 5 minutes long, select a recognition language and, if necessary, add text explanations. After pressing the 'Generate' button, the audio is processed locally, without sending to external servers. The result appears in a separate panel as text with timestamps, which can be copied or downloaded in SRT format.
Important to understand: the tool does not perform automatic ideal subtitle splitting — it only recognizes speech and sets timestamp boundaries. Quality depends on the clarity of speech, background noise and the selected language. If the audio contains multiple speakers or specialized terminology, the result may require manual editing.
The interface has no sliders or style presets. Everything you can change is the language and the text field for additional instructions (up to 1000 characters). This means that to get different variants you need to vary exactly these parameters.
Upload an audio file, select the language 'Russian' and click 'Generate'. After a few seconds you will get an SRT file with subtitles.
Which parameters can be changed to get different variants
The first and most obvious parameter is the recognition language. If the audio contains mixed speech (for example, Russian with English terms), you can try selecting different languages. This will give two subtitle variants that will differ at the code-switching points.
The second parameter is the text field for additional instructions. Here you can enter context: conversation topic, proper names, specific terms. For example, if you know that 'Ivan Petrov' and 'project Alpha' are mentioned in the audio, indicate this — the tool may recognize these words more accurately. You can also ask to preserve punctuation or split into short lines, although the latter is not guaranteed.
The third way is preliminary audio processing. If the recording is noisy, you can first clean it up in another editor, and then upload to 'Subtitles'. This is not a function of the tool itself, but a perfectly acceptable external step. The cleaner the sound, the more accurate the recognition.
For a biology lecture audio you can specify in additional instructions: 'Topic: photosynthesis. Terms mentioned: chloroplast, thylakoid, stroma'. This increases the chances of correct spelling.
Comparing three subtitle variants
Suppose you have an interview audio recording lasting 3 minutes. You create three variants: the first — with the language 'Russian' and no additional instructions; the second — with the language 'Russian' and specifying names and terms; the third — with the language 'English' (in case there are many anglicisms in the speech). Compare them by several criteria: word recognition accuracy, correct timestamp placement and readability convenience.
Word accuracy is easily checked by reading the text and marking errors. Timestamps can be evaluated by opening the SRT in any player and comparing synchronization. Readability convenience depends on string length: if subtitles are too long, they are hard to read. The tool does not guarantee ideal length, but sometimes additional instructions help.
Do not expect that one variant will be ideal. Often the best result is a combination: for example, take accurate timestamps from the first variant and correct names from the second. Manual editing will still be required, but the better the initial variants, the less work.
Variant 1: 85% of words are correct, but the name 'Alexey' is recognized as 'Alexander'. Variant 2: 92% of words are correct, name fixed. Variant 3: 70% of words are correct, but timestamps are off. Choose variant 2 as the main one.
How to choose the best variant and refine it
After comparison choose the variant with the fewest word errors and most accurate timestamps. If there are errors, they can be manually corrected in any text editor. SRT is a simple text format where each subtitle has a number, a timestamp interval and text. The main thing is not to break the structure: number, then time in format HH:MM:SS,mmm --> HH:MM:SS,mmm, then text.
When refining pay attention to string length. It is recommended that one subtitle line contain no more than 42 characters and no more than two lines on screen. If the tool created too long lines, split them manually. Also check that timestamps do not overlap and are not too short (less than 1 second).
If you are not sure about the correctness of some fragment, you can go back to the audio and listen to it. The tool does not allow editing subtitles within the interface, so all edits are made in an external editor. Save the file in UTF-8 encoding to avoid problems with Cyrillic display.
After choosing variant 2 you open the SRT in a text editor, fix 'Alexander' to 'Alexey', split a long line into two and save the file.
Practical example: subtitles for a short video
Imagine you recorded a video greeting lasting 2 minutes. You want to get subtitles. Upload the audio to the tool, select the language 'Russian', in additional instructions write: 'Topic: new product announcement. Mentions names: SmartWatch X, Health app'. Press 'Generate' and get SRT.
Then you decide to make a second variant: same settings, but without additional instructions. Compare: in the first variant product names are written correctly, in the second — with errors. Timestamps in both variants are roughly the same. You choose the first variant and manually correct a few minor errors.
This example shows that even simple addition of context can improve the result. However do not overload instructions: they are limited to 1000 characters, and too general phrases may not help. Better to specify specific names and terms.
Instruction: 'Topic: new product announcement. Mentions names: SmartWatch X, Health app'. Result: names recognized correctly.
Tool limitations and typical errors
The tool does not support batch processing, automatic saving or integration with other services. You cannot upload multiple files at once or get subtitles in DOCX or PDF format. Export is possible only in TXT (without timestamps) and SRT (with timestamps). Also there is no possibility to edit subtitles within the interface — only copy or download.
Recognition may fail on proper names, abbreviations, special terms and with strong background noise. Timestamps sometimes shift, especially if the audio has pauses or multiple speakers. The tool does not identify speakers and does not separate their replicas.
Do not expect an ideal result without manual editing. Always check subtitles for correspondence to the audio. If audio is longer than 5 minutes or more than 20 MB, the tool will not accept the file. In such cases you need to split the audio into parts and process separately.
When uploading a file of 25 MB the tool will give an error. You will need to compress the audio or split it into parts.
How to check subtitle quality
The first step is visual review of the SRT in a text editor. Check that each block starts with a number, then comes a timestamp, then text. Make sure there are no empty blocks or sequence violations.
The second step is synchronization with video. Open a video player that supports subtitle loading (e.g. VLC) and enable SRT. Review several fragments: do subtitles match the speech. If there are discrepancies, write down the timecodes and fix manually.
The third step is spelling and punctuation check. The tool may skip commas or capital letters. Run the text through any text editor with spell checking. This will take a little time but significantly improve readability.
In VLC subtitles appeared with a 0.5 second delay. You open the SRT and shift all timestamps 0.5 seconds earlier.
The 'Subtitles' tool is a simple way to get an SRT file with timestamps, working locally in the browser. To get different variants, change the recognition language and add context to the text field. Compare variants by word accuracy and timestamps, then refine the best one manually. Remember the limitations: no batch processing, no editing within the interface and no guarantees of perfect recognition. Always check the result and, if necessary, clean up the audio before uploading. Following these tips, you can efficiently create subtitles for your videos.
Open tool: Subtitle generator ↗Interface translations are machine-assisted and may need refinement.